--- pretty_name: J dataset_info: - config_name: Github_easy features: - name: json_schema dtype: string - name: unique_id dtype: string splits: - name: train num_bytes: 1208636 num_examples: 1170 - name: val num_bytes: 182688 num_examples: 191 - name: test num_bytes: 539656.0 num_examples: 577 download_size: 540610 dataset_size: 1930980.0 - config_name: Github_hard features: - name: json_schema dtype: string - name: unique_id dtype: string splits: - name: train num_bytes: 12816152 num_examples: 746 - name: val num_bytes: 1607525 num_examples: 122 - name: test num_bytes: 5754647.483870967 num_examples: 368 download_size: 3562146 dataset_size: 20178324.48387097 - config_name: Github_medium features: - name: json_schema dtype: string - name: unique_id dtype: string splits: - name: train num_bytes: 4990832 num_examples: 1189 - name: val num_bytes: 557390 num_examples: 194 - name: test num_bytes: 2417201.5784148397 num_examples: 586 download_size: 1580336 dataset_size: 7965423.57841484 - config_name: Github_trivial features: - name: json_schema dtype: string - name: unique_id dtype: string splits: - name: train num_bytes: 467333.24324324325 num_examples: 266 - name: val num_bytes: 77303.24324324324 num_examples: 44 - name: test num_bytes: 235423.51351351352 num_examples: 134 download_size: 158044 dataset_size: 780060.0 - config_name: Github_ultra features: - name: json_schema dtype: string - name: unique_id dtype: string splits: - name: train num_bytes: 7311744.743902439 num_examples: 98 - name: val num_bytes: 1193754.243902439 num_examples: 16 - name: test num_bytes: 3730482.012195122 num_examples: 50 download_size: 2221455 dataset_size: 12235981.0 - config_name: Glaiveai2K features: - name: json_schema dtype: string - name: unique_id dtype: string splits: - name: train num_bytes: 865943.3989455184 num_examples: 1026 - name: val num_bytes: 141791.9015817223 num_examples: 168 - name: test num_bytes: 432971.6994727592 num_examples: 513 download_size: 284264 dataset_size: 1440707.0 - config_name: JsonSchemaStore features: - name: json_schema dtype: string - name: unique_id dtype: string splits: - name: train num_bytes: 13308367.977642277 num_examples: 295 - name: val num_bytes: 2210542.4776422763 num_examples: 49 - name: test num_bytes: 6676740.544715447 num_examples: 148 download_size: 4019966 dataset_size: 22195651.0 - config_name: Kubernetes features: - name: json_schema dtype: string - name: unique_id dtype: string splits: - name: train num_bytes: 15388503.69924812 num_examples: 639 - name: val num_bytes: 2528627.3684210526 num_examples: 105 - name: test num_bytes: 7706292.932330827 num_examples: 320 download_size: 6819424 dataset_size: 25623424.0 - config_name: Snowplow features: - name: json_schema dtype: string - name: unique_id dtype: string splits: - name: train num_bytes: 969083.2952853598 num_examples: 242 - name: val num_bytes: 160179.0570719603 num_examples: 40 - name: test num_bytes: 484541.6476426799 num_examples: 121 download_size: 298277 dataset_size: 1613804.0 - config_name: WashingtonPost features: - name: json_schema dtype: string - name: unique_id dtype: string splits: - name: train num_bytes: 1604526.016 num_examples: 74 - name: val num_bytes: 281876.192 num_examples: 13 - name: test num_bytes: 823945.792 num_examples: 38 download_size: 565170 dataset_size: 2710348.0 - config_name: default features: - name: json_schema dtype: string - name: unique_id dtype: string splits: - name: train num_bytes: 54520620 num_examples: 5754 - name: val num_bytes: 15255546 num_examples: 937 - name: test num_bytes: 27031812.394351464 num_examples: 2867 download_size: 20765998 dataset_size: 96807978.39435147 configs: - config_name: Github_easy data_files: - split: train path: Github_easy/train-* - split: val path: Github_easy/val-* - split: test path: Github_easy/test-* - config_name: Github_hard data_files: - split: train path: Github_hard/train-* - split: val path: Github_hard/val-* - split: test path: Github_hard/test-* - config_name: Github_medium data_files: - split: train path: Github_medium/train-* - split: val path: Github_medium/val-* - split: test path: Github_medium/test-* - config_name: Github_trivial data_files: - split: train path: Github_trivial/train-* - split: val path: Github_trivial/val-* - split: test path: Github_trivial/test-* - config_name: Github_ultra data_files: - split: train path: Github_ultra/train-* - split: val path: Github_ultra/val-* - split: test path: Github_ultra/test-* - config_name: Glaiveai2K data_files: - split: train path: Glaiveai2K/train-* - split: val path: Glaiveai2K/val-* - split: test path: Glaiveai2K/test-* - config_name: JsonSchemaStore data_files: - split: train path: JsonSchemaStore/train-* - split: val path: JsonSchemaStore/val-* - split: test path: JsonSchemaStore/test-* - config_name: Kubernetes data_files: - split: train path: Kubernetes/train-* - split: val path: Kubernetes/val-* - split: test path: Kubernetes/test-* - config_name: Snowplow data_files: - split: train path: Snowplow/train-* - split: val path: Snowplow/val-* - split: test path: Snowplow/test-* - config_name: WashingtonPost data_files: - split: train path: WashingtonPost/train-* - split: val path: WashingtonPost/val-* - split: test path: WashingtonPost/test-* - config_name: default data_files: - split: train path: data/train-* - split: val path: data/val-* - split: test path: data/test-* license: mit task_categories: - text-generation --- # JSONSchemaBench [![Paper](https://img.shields.io/badge/Paper-arXiv-blue)](https://arxiv.org/abs/2501.10868) [![GitHub](https://img.shields.io/badge/Code-GitHub-blue)](https://github.com/guidance-ai/jsonschemabench) JSONSchemaBench is a benchmark of **real-world JSON schemas** designed to evaluate **structured output generation** for Large Language Models (LLMs). It contains approximately **10,000 JSON schemas**, capturing diverse constraints and complexities. ```python import datasets from datasets import load_dataset def main(): # Inspect the available subsets of the dataset all_subsets = datasets.get_dataset_config_names("epfl-dlab/JSONSchemaBench") print("Available subsets:", all_subsets) # Example output: ['Github_easy', 'Github_hard', 'Github_medium', 'Github_trivial', 'Github_ultra', 'Glaiveai2K', 'JsonSchemaStore', 'Kubernetes', 'Snowplow', 'WashingtonPost', 'default'] # Access a specific subset of the dataset subset_name = "Github_easy" github_easy = load_dataset("epfl-dlab/JSONSchemaBench", subset_name) print(f"Loaded subset '{subset_name}':", github_easy) # Load the entire dataset as a whole entire_dataset = load_dataset("epfl-dlab/JSONSchemaBench", "default") print("Loaded entire dataset:", entire_dataset) if __name__ == "__main__": main() ``` ## Update (March 31st, 2025) To improve inference efficiency and streamline data collation, we’ve decided to drop a small number of exceptionally long samples from the dataset. We’re using the `meta-llama/Llama-3.2-1B-instruct` tokenizer, and the filtering criteria are as follows: - Github_easy: Samples longer than 1024 tokens — 5 out of 582 removed - Github_medium: Samples longer than 2048 tokens — 7 out of 593 removed - Github_hard: Samples longer than 8192 tokens — 4 out of 372 removed - Other subsets are not touched Since the number of discarded samples is minimal, this change is expected to have at most a 1% impact on results. ## ⚠️ Important Update (March 10th, 2025) We have restructured the dataset to include train/val/test splits. If you downloaded the dataset before this date, you might encounter errors like `KeyError: 'Github_easy'`. To fix this issue, please follow one of the options below: 1. Update How Subsets Are Accessed: If you previously used: ```python from datasets import load_dataset, concatenate_datasets, DatasetDict, Dataset subset: DatasetDict = load_dataset("epfl-dlab/JSONSchemaBench") subset["Github_easy"] ``` You can update it to: ```python from datasets import load_dataset, concatenate_datasets, DatasetDict, Dataset subset: DatasetDict = load_dataset("epfl-dlab/JSONSchemaBench", name="Github_easy") subset: Dataset = concatenate_datasets([subset["train"], subset["val"], subset["test"]]) ``` 2. Load the Dataset in the Old Structure: If you need the previous structure, you can use a specific revision: ```python dataset = load_dataset("epfl-dlab/JSONSchemaBench", revision="e2ee5fdba65657c60d3a24b321172eb7141f8d73") ``` We apologize for the inconvenience and appreciate your understanding! 😊 ## 📌 Dataset Overview - **Purpose:** Evaluate the **efficiency** and **coverage** of structured output generation. - **Sources:** GitHub, Kubernetes, API specifications, curated collections. - **Schemas:** Categorized based on complexity and domain. ### 📊 Dataset Breakdown | Dataset | Category | Count | | --------------- | ------------------- | ----- | | GlaiveAI-2K | Function Call | 1707 | | Github-Trivial | Misc | 444 | | Github-Easy | Misc | 1943 | | Snowplow | Operational API | 403 | | Github-Medium | Misc | 1976 | | Kubernetes | Kubernetes API | 1064 | | Washington Post | Resource Access API | 125 | | Github-Hard | Misc | 1240 | | JSONSchemaStore | Misc | 492 | | Github-Ultra | Misc | 164 | | **Total** | | 9558 | ## 📥 Loading the Dataset ```python from datasets import load_dataset dataset = load_dataset("epfl-dlab/JSONSchemaBench") print(dataset) ``` ## 🔍 Data Structure Each dataset split contains: - `"json_schema"`: The schema definition. - `"unique_id"`: A unique identifier for the schema. 🚀 **For more details, check out the [paper](https://arxiv.org/abs/2501.10868).** ## 📚 Citation ```bibtex @misc{geng2025jsonschemabench, title={Generating Structured Outputs from Language Models: Benchmark and Studies}, author={Saibo Geng et al.}, year={2025}, eprint={2501.10868}, archivePrefix={arXiv}, primaryClass={cs.CL}, url={https://arxiv.org/abs/2501.10868} } ``` ## License This dataset is provided under the [MIT License](https://opensource.org/licenses/MIT). Please ensure that you comply with the license terms when using or distributing this dataset. ## Acknowledgements We would like to thank the contributors and maintainers of the JSON schema projects and the open-source community for their invaluable work and support.